MLOC mangyshlak cluster

Example Cluster: Mangyshlak PNEs

The Mangyshlak cluster illustrates indirect calibration using a set of three Peaceful Nuclear Explosions (PNEs) conducted by the U.S.S.R. in western Kazakhstan, termed the “Mangyshlak” sequence by the Russians. Although nuclear tests are often thought of as “ground truth” the record-keeping for PNEs was not done to the standards typical of a major nuclear test site and there is still quite a bit of uncertainty concerning the precise hypocentral coordinates of many PNEs. Indirect calibration was done in this case as a consistency check on the reported epicenters and shot times of these three explosions. Details of this study can be found in Mackey and Bergman (2014).

The series included in the distribution is mangyshlak4. This cluster has already undergone the cleaning process in previous series of runs, so the command file mangyshlak4.1.cfil takes empirical reading errors and the observed spread of various travel-time branches from the previous run (commands rfil and tfil):

rfil mangyshlak3.5.rderr
tfil mangyshlak3.5.ttsprd

The two output files from mangyshlak3.5 are included in the mangyshlak4 directory.The full command file for mangyshlak4.1 is shown here:

pltt 1 2 3 6
ccat mangyshlak_summary.txt
rfil mangyshlak3.5.rderr
tfil mangyshlak3.5.ttsprd
hlim 30. 90.
clim 0. 180.
wind 3 4
frec 1 1 0 1
freh 1 1 0 1
comm elevation in the source area is 200 m
memb
even 19691206.0702.59
inpu 19691206.0702.59.mnf
depe 0.2 0.1
comm OT and depth from Sultanov
calf 7 2 59.85 43.867  54.800  0.2 0.0 0.01 0.01 0.04 1.0 ! Sultanov geodetic
calf 7 2 59.85 43.8625 54.7727 0.2 0.0 0.01 0.01 0.04 1.0 ! Mackey site visit
memb
even 19701212.0700.59
inpu 19701212.0700.59.mnf
depe 0.3 0.1
comm OT and depth from Sultanov
calf 7 0 59.83 43.85   54.80   0.3 0.0 0.01 0.01 0.04 1.0 ! Sultanov seismic
calf 7 0 59.83 43.9096 54.7937 0.3 0.0 0.01 0.01 0.04 1.0 ! Mackey site visit
memb
even 19701223.0700.59
inpu 19701223.0700.59.mnf
depe 0.3 0.1
comm OT and depth from Sultanov
calf 7 0 59.76 44.025  54.993  0.3 0.0 0.01 0.01 0.04 1.0 ! Sultanov geodetic
calf 7 0 59.76 43.8858 54.8973 0.3 0.0 0.01 0.01 0.04 1.0 ! Mackey site visit

The presence of two calf commands (a variation meaning the focal depth is calibrated but not the origin time) for each event is notable, and it leads to a warning from mloc during the run that there are multiple calibration commands for the event:

Enter commands:

: cfil mangyshlak4.1.cfil
  FYI from proc_cal_: event   1 19691206.0702.59 is declared as a calibration event
  FYI from proc_cal_: event   1 19691206.0702.59 has multiple calibration commands
  FYI from proc_cal_: event   2 19701212.0700.59 is declared as a calibration event
  FYI from proc_cal_: event   2 19701212.0700.59 has multiple calibration commands
  FYI from proc_cal_: event   3 19701223.0700.59 is declared as a calibration event
  FYI from proc_cal_: event   3 19701223.0700.59 has multiple calibration commands

mloc will use the last calibration command found for an event, so the warning should simply be taken as a reminder that the user should be sure that the order of commands is as intended. In this case the first calibration hypocenter is from a publication that attempted to compile the most reliable information for all PNEs. The second calibration location (epicenter) is the one measured with GPS by Kevin Mackey at the actual boreholes during a site visit in 2013, guided by the results of the mloc analysis.

Nearly all commands in mloc can be issued more than once in a command file, and the last instance will take effect. This is quite useful with dep_ commands, for there are often multiple estimates of depth for an event.

Another feature to note is the use of in-line comments (beginning with “!”) in the “calf” commands to identify the source of the calibration data. In-line comments can be used with any command.

Even though the residual shift vectors (remaining errors in location after the calibration shift) are small by most standards, ranging from 250-600 m, inspection of the ~.cal file for mangyshlak4.1 reveals that the consistency of the relative locations determined by mloc is a bit marginal when compared to the calculated uncertainties. For example here is the first estimate of radius of doubt:

Radius of doubt test based on null hypothesis that all residual cluster vectors have zero length
Critical value =      5.2 at 90%

Individual event contributions to kocs2:
  1     4.28889
  2     0.44913
  3     0.12304
rdbt_test =   0.00; observed value =   4.86; null hypothesis cannot be rejected    
rdbt1 =   0.00

Although the radius of doubt comes out zero, it is a close thing; the observed value of the test statistic (4.86) is almost as large as the threshold (5.2 at 90% confidence) for rejecting the null hypothesis that all residual shift vectors have zero length. The second test, based on “coverage” of the confidence ellipses indicates that the radius of doubt could be non-zero:

Radius of doubt based on coverage statistics

Threshold percentage of probability for radius of doubt test: 0.10
                  rdbt    nr    nrc    k    coverage   P(X  k)
rdbt_test =      0.000     3     1     2     0.333     0.028; null hypothesis is rejected
rdbt_test =      0.200     3     1     2     0.333     0.028; null hypothesis is rejected
rdbt_test =      0.400     3     2     1     0.667     0.271; null hypothesis cannot be rejected
rdbt2 =   0.40

This test is not actually appropriate in this case because of the number of data. It requires at least 10 calibration events to be applicable. For any number less there would need to be complete coverage in order not to reject the null hypothesis.

All statistical arguments in this case are on shaky ground because with only three events the power of our robust estimator to accurately estimate the empirical reading errors is rather limited.

The final aspect of the mangyshlak cluster to note here concerns the use of the ccat command, used to create a special pair of output files designed for easy import into the GCCEL database. These files are stored in the directory mangyshlak4.1_comcat. One, mangyshlak4.1_plots.pdf contains all the plots that were requested and the other, mangyshlak4.1.comcat is a text file containing all data and results in an MNF v1.4-formatted bulletin.

The argument to the ccat command is the name of a text file that is treated as a commentary on the cluster; it is inserted near the top of the ~.comcat file. The text file can have any name; here it is named mangyshlak_summary.txt. It must be hard wrapped at some reasonable line length.

Last Updated on